In a separate post, I wrote about EV charging as a utilization problem embedded in real estate. This essay zooms out to the market structure around that world: why public charging networks have torched so much capital, why software can't capture much value on its own, and why the gravity of the business pulls toward vertical integration.

The charging infrastructure market is racing toward a $100B+ opportunity by 2040, but the thesis that pure-play software companies will dominate misreads the fundamental economics. Value is accruing to vertically integrated operators who control real estate, customer relationships, and software, not to asset-light platforms or infrastructure installers. Tesla remains the only profitable charging network, and the greatest returns will flow to retail giants like Walmart and automaker consortiums who can subsidize infrastructure with adjacent revenue streams. Pure software plays succeed only as B2B enablers, not consumer-facing winners.

This essay is my attempt to lay out that thesis and the market map behind it. I've tried to make it accessible to readers without deep EV industry or financial knowledge while keeping enough detail for those who want to test the logic. If you disagree with anything here, especially the investment implications, I'd genuinely welcome the pushback.

The working hypothesis that "winners will look like software/coordination businesses, not infrastructure installers" contains a critical truth but misses the endgame: software is necessary but not sufficient. The real winners control the full stack.

Business Model Crisis

The SPAC boom^* delivered three major public charging networks, ChargePoint (CHPT), EVgo (EVGO), and Blink (BLNK). All three have been catastrophically unprofitable despite explosive revenue growth. ChargePoint lost $253M on $417M revenue in fiscal 2025, EVgo burned $127M on $257M, and Blink lost $198M on $126M. Their stocks have collapsed roughly 82%^**, 85%, and 98% respectively from their 2021 peaks.

*A SPAC is a shortcut to going public: investors create an empty shell company, take it public, then merge it with a private company. In 2020-2021, low interest rates and speculative hype created a frenzy where SPACs took hundreds of companies public, often with billion-dollar valuations based on optimistic projections.

**ChargePoint did a reverse stock split in 2025, so the price comparisons were adjusted for that. Data shows an adjusted all-time high of $922 (post-split adjusted), but the unadjusted peak was ~$49. The key point remains: these companies have destroyed massive amounts of shareholder value since the SPAC-era highs.

Tritium, a major DC fast charger manufacturer founded in 2001 that went public via SPAC in 2022, filed for insolvency in April 2024 after never achieving profitability, hemorrhaging $121M on $185M revenue in its final fiscal year. Electric bus and battery maker Proterra, which also went public via SPAC in 2021, filed Chapter 11 in August 2023. Charging infrastructure provider Charge Enterprises followed with its own Chapter 11^* filing in March 2024.

*Chapter 11 is a U.S. bankruptcy process that lets a company hit "pause" on its debts and keep operating while it restructures, rather than immediately selling off everything. The business can survive, but shareholders usually lose everything.

Tesla Supercharger stands alone as the profitable exception. While Tesla doesn't break out Supercharger financials, its "Services and Other" category, which includes charging, grew profit 64% sequentially in Q2 2025, with Bloomberg estimating the Supercharger network will generate $7.4B revenue at 10% margins by 2030. The network operates 70,000+ connectors globally with 97% uptime.

Unit Economics Explain Why Infrastructure Alone Fails

The brutal math of DC fast charging reveals why pure infrastructure plays struggle. A typical 4-charger, 150kW station in California costs $600-800K in capex. At current average utilization of 16.1%, these sites generate $265-285K annually.

McKinsey's analysis shows break-even requires 20% utilization without subsidies or 15% with incentives, but national averages sit at 16.1% as of Q2 2025. High-performing metro areas like Las Vegas hit 45.5% peak utilization, while rural states like South Dakota limp along at 1.1%. The geographic dispersion creates structural unprofitability. Networks must serve low-density areas to provide useful coverage, even if those sites never achieve viable economics.

The capital intensity dwarfs comparable businesses. DC fast chargers cost $300K installed versus $20K for gas pumps. Yet the margin on selling marked-up electricity is razor-thin, and 80% of EV charging happens at home, severely limiting the addressable market.

EVgo, which actually owns and operates its 4,590 charging stalls, is approaching break-even with 35% charging network gross margins and expects positive adjusted EBITDA in Q4 2025. But even EVgo's progress depends heavily on subsidies (which are set to expire soon).

Software Creates Value but Struggles to Capture it

Watching infrastructure players bleed cash, investors pivoted to a cleaner thesis: forget owning chargers, build the operating system. Software platforms could aggregate demand across fragmented hardware, optimize pricing and utilization, and handle payments and customer support, (all the things ChargePoint and EVgo were failing at) all without the capital intensity.

The thesis was right in that the incumbents were embarrassingly bad at the basics. With broken chargers, payment failures, and zero customer support, user experience was genuinely a nightmare. Good software creates value by fixing these problems. But that value flows to whoever controls the real estate and customer relationship, not to the middleware provider charging 15% transaction fees.

The economics turned toxic with the rise of OCPP^*. A property owner can swap networks without replacing a single piece of hardware. Strong utilization lets you charge premium fees. If utilization is weak, networks can still earn some baseline revenue per site, but churn risk^* rises and the platform's reputation may be harmed.

*OCPP (Open Charge Point Protocol) is an open communication standard that allows EV chargers to interface with any compatible network management system, regardless of hardware manufacturer. For network providers, OCPP compliance makes initial sales easier (site hosts aren't locked into one ecosystem), but also eliminates switching costs.

The path to scale makes it worse. Unlike consumer, you can't grow virally. Even when you win, you're trapped in a low-margin, high-churn business where the customer has all the leverage.

The market delivered its verdict quickly. Driivz, once positioned as the "operating system for EV charging," got acquired by Vontier for parts, not spun out as the software giant investors imagined. The real money went to companies that controlled both the hardware and the site relationship, not the ones trying to be neutral platforms in between.

Real Estate Hosts Treat Charging as an Amenity

Most property owners don't think of EV charging as a standalone business. They treat it like better lighting or nicer landscaping: something that makes the asset more attractive, keeps tenants happy, and helps fill parking lots with the "right" kind of customer.

In the most common structure today, a third-party operator fronts the capex, handles interconnection, owns the hardware, and runs the network. The host kicks in parking spaces, some electrical capacity, then gets a small rev share, fixed ground lease, or just the intangible benefit of being "EV friendly."

On a per-site basis, the absolute dollars a landlord can earn from charger revenue are tiny relative to rent roll or in-store sales. What matters is traffic quality and dwell time. If chargers help fill units faster, push renewal rates higher, or grow basket size inside the building, they're doing their job, even if the station barely breaks even itself.

We've already seen this movie once with fuel. Gas became a low-margin commodity, and the real profits migrated past the doorstep. EV charging is running that playbook in fast-forward, with the winners being the people who own the doors.

Walmart and Retailers will Dominate

If the pure-play charging networks are failing and software can't capture value independently, where does the smart money flow? Follow Walmart, which plans to install EV charging at "thousands" of locations by 2030, potentially doubling the U.S. DC fast charger count.

Walmart's model inverts the charging economics, similar to what we were talking about earlier: infrastructure serves as a traffic driver and dwell time monetization play, not a profit center. The target 20-30 minute charging window aligns perfectly with shopping trips. The company uses charging as a loss leader while capturing margin on in-store purchases during extended dwell time. Walmart+ members receive additional discounts, creating subscription service attachment.

This mirrors the century-old gas station evolution: fuel became a razor-thin margin commodity with station owners earning just 7 cents per dollar spent on gas, while convenience stores generate 30-35% gross margins and deliver the majority of net profits. Walmart's scale, existing infrastructure, and integrated business model can tolerate charging margins that would bankrupt standalone operators.

Target follows with 100+ locations planned across 20+ states, partnering with Tesla, ChargePoint, and Electrify America. The automaker consortium IONNA^* committed $250M over three years for 1,000 charging bays in California alone, targeting 30,000 bays nationwide by 2030.

*IONNA is a joint venture of eight major automakers (BMW, GM, Honda, Hyundai, Kia, Mercedes-Benz, Stellantis, and Toyota) formed to build a North American charging network. Unlike traditional charging networks, IONNA stations are designed as destination charging hubs with amenities, retail integration, and Plug & Charge technology that automatically handles payment. For the automakers, this is defensive infrastructure spending: vehicle sales depend on charging confidence, and they can't rely on third parties to deliver reliable customer experience.

Oil majors view charging as a customer retention strategy as gasoline volumes decline, leveraging existing real estate and convenience store expertise. They can afford to operate charging at or below breakeven to preserve location traffic.

Vertical Integration is the Market Structure

By 2030, I see the U.S. DC fast charging market consolidating from 50-plus players down to roughly 8-12 networks controlling about 80% of public fast charging. These leaders will all be entities that control real estate, balance sheets, and a primary profit center that is not the charger itself.

Cell towers started as a fragmented mess of regional operators and carrier-owned sites in the 1990s and ended with three REITs that control around three-quarters of U.S. macro towers. Gas stations peaked at a little over 200,000 sites in the late 20th century and then bled out thousands of small independents as environmental regulations and razor-thin fuel margins pushed the industry into the arms of big chains and c-stores. Telecom companies in the late 1990s poured more than $500 billion into fiber-optic and related network infrastructure, went bankrupt en masse, and then quietly made private equity investors rich a decade later once demand finally caught up.

EV charging sits on the same curve, just running on a slightly faster clock. 2020-2026 will be the overbuild and idealism phase. 2026-2030 will be the profitability inflection phase, when average utilization climbs into a range where stations actually consistently generate profit. And finally, 2030-2035 will be the oligopoly phase, where the surviving networks have enough density, enough capital, and enough adjacent revenue streams that the game shifts away from charger spam.

Capital Intensity versus Software Economics

Going back to the unit analysis, we mentioned how a typical 4-charger, 150kW station would cost anywhere between $600K and $800K in capex.

Now look at the software slice of that stack. Most charging management platforms charge $30-$100 per port per month. On a 4-port site, that's $1,440-$4,800 a year in recurring revenue sitting on top of a half-million-dollar asset. You're capturing maybe 0.3-1.0% of the capital you supposedly "operate." The rev-share variants don't change the physics much, they tax sessions instead of ports, but they're still skimming a small percentage of gross revenue without owning the thing that sets volume or price.

That's the real weakness in the "we'll be the operating system of EV charging" story. The people writing the $400,000 checks can switch between network providers with essentially no hardware replacement cost.

Policy risk just amplifies it. As we know, the 30C tax credit can swing a site's projection from deeply negative to comfortably positive. When that incentive effectively sunsets for a big chunk of use cases in 2026, a lot of marginal stations that looked fine in underwriting models won't look too great anymore. The only actors who can ride that out are the ones who can make the loss back somewhere else [retail baskets, you get where I am going here].

A software-only platform doesn't have that option. Most of the time, they don't have any say where the concrete gets poured, what the power costs, how pricing is set, or whether the landlord dumps you for a cheaper competitor once your contract is up.

Where Software Still Matters (and where it doesn't)

This doesn't mean software is irrelevant. Good software is the only reason this ecosystem functions at all. Somebody has to route drivers, manage queues, arbitrate between onsite loads, optimize time-of-use rates, handle payments and refunds, and interface with a dozen different utilities and incentive programs.

There are two slices of the stack where I think software looks genuinely interesting.

The first is pre-construction: site selection and underwriting. Deciding which retail center to put hardware at is a data problem as much as it is an electrical one. If you can consistently pick above-average sites, you create real value. Do that at scale and your model gets better over time as you observe actual utilization against your forecasts (Stable Auto has done this well).

The second is bundling: making yourself sticky by wrapping software with things that are not easily commoditized. That might mean locking in multi-year operations and maintenance contracts, offering financing so hosts don't have to write big checks out of pocket, or stitching together enough volume to negotiate better hardware pricing than a single landlord ever could. The moment you start doing that, though, you edge away from pure software and toward being a developer or operator.

What almost certainly won't work at scale is the idea of being a neutral, asset-light consumer brand that "owns" the driver relationship while outsourcing everything else. Drivers will go where the chargers are. Hosts will work with whoever can get projects financed and built. The small gap in between those two realities is not a big enough space to build an enduring, high-margin consumer software business.

The coming 24 months will be decisive as incentives expire and true private-market economics emerge. The winners will control distribution, customer relationships, and software, but not any single layer independently. This is a capital-intensive infrastructure business.

Software is necessary but not sufficient. In a world of expensive boxes bolted to concrete, the center of gravity sits with whoever is willing to own the concrete, and wait.

If this essay is the outside view, its pair is the inside view. If you haven't already read it, the first part stays near to the property line, following how a single multifamily building or retail center actually decides whether to add chargers, how reservations change utilization, and what it looks like to align incentives between owners, capital providers, networks, and drivers. Read together, the two posts are the same argument from opposite directions. Utilization, not installation, is what makes infrastructure real, and the value settles with whoever can both influence that utilization and afford to be patient.